SOURCE CONSIDERATIONS FOR MODERATE MAGNITUDE EARTHQUAKE GROUND MOTION SIMULATION VALIDATION - ROBIN LEE1, BRENDON BRADLEY1, ROBERT GRAVES2

Source Considerations for Moderate
Magnitude Earthquake Ground Motion
 Simulation Validation
 Robin Lee1, Brendon Bradley1, Robert Graves2
 1University of Canterbury and QuakeCoRE, 2USGS Pasadena

 1

Motivation
• Ground motion simulation validation allows us to use observations to
 quantify the predictive capability and infer where improvements can
 be made.
• Previously focussed on large Mw and small Mw.
• Large Mw studies on Darfield, Christchurch and Kaikōura earthquakes
 (Razafindrakoto et al. (2018) and Bradley et al. (2017)).
• Small Mw (3.5-5.0) studies for Canterbury and NZ-wide (Lee et al.
 (2020) and not yet published).
• Moderate (5.0-7.0) Mw currently missing.
• Moving from small to moderate, there are several additional
 considerations required.
 2

NZ Validation Progression
 2015 – 2016 2017 – 2018 2018 – 2019 2020 – ?
 Mw 7.1 Darfield
Events Mw 6.2 Christchurch 148 Small Mw 498 Small Mw 62 Moderate Mw
 Mw 7.8 Kaikōura

Region Source-specific Canterbury New Zealand “New Zealand”

 NZVM v1.66 NZVM v1.66 NZVM v2.02 NZVM v2.03
Velocity
Model (Canterbury Basin (Canterbury Basin (8 sedimentary (9 sedimentary
 only) only) basins) basins)

Sim Standard Graves Standard Graves Modified Graves Modified Graves
Method and Pitarka and Pitarka and Pitarka and Pitarka
 (2010,2015) (2010,2015) (2010,2015) (2010,2015,2016)
 3

Simulation Methodology
• Widely-used Graves and Pitarka (GP)
 hybrid approach with modifications.
• Low frequency (LF) component (f0.5Hz)
 from simplified physics-based wave
 propagation.
• Period-dependent empirical Vs30-based
 site amplification (HF only).
• LF and HF merged to produce broadband
 (BB) ground motion.
 4

Earthquake Events

 Small Moderate
Magnitude Magnitude

 5

Earthquake Events
• 62 earthquakes.
• 1738 high quality records across
 203 stations from an initial set of
 over 4000.
• Quality of records determined
 from neural network trained
 against small Mw earthquakes.
• Enforced 3 HQ records per event
 and per station.
• Largest is Mw 6.6 which is 2013
 Lake Grassmere EQ.
 6

Earthquake Events
• Limited to events with centroid moment tensor solutions with
 centroid depth ≤ 20km.
• Larger Mw on average more records.

 7

Recording Stations
• Broadband and strong motion stations across NZ.
• Vs30 values from Foster et al. (2019) NZ-wide Vs30 map.
• A variety of site conditions as shown by range of Vs30.
• Previous validation showed many sites appeared to have Vs30 too low.
• Model currently being modified.

 8

Velocity Modelling
 Waikato/

 9

Velocity Modelling

 10

Source Modelling
• Sources for small Mw can be reasonably approximated as point
 sources.
• Not reasonable for moderate Mw.
• Need to model as finite fault.
• Additional complexities.
 • Rupture geometry.
 • Slip, rise time and rake distribution.
 • Temporal evolution of slip.
 • Hypocentre location (currently set as centre of rupture plane).

 11

Rupture Geometry
• Source description from geonet centroid moment tensor catalogue.
• To get a finite fault we use Mw-Area scaling relationship.
• Use Leonard (2010).

 12

Magnitude-Area Scaling Relationship
• Predicted A does not equal product of predicted L and W.
• I won’t rule out an error or misunderstanding on my end but I
 checked this very rigorously. Has anyone else checked this before?

 13

Magnitude-Area Scaling Relationship
• We consider two options:
1. L and W are equal so L = W = 
2. L and W are have a ratio r=L/W equal that predicted by Leonard
 (2010) equations while preserving the A predicted by Leonard.
 • Example:
 • For a Mw 5.0 strike slip earthquake:
 • Leonard gives us A = 10.2, L = 3.1, W = 2.8
 • L x W = 3.1 x 2.8 = 8.8 ≠ 10.2
 • Instead use L = 3.4 and W = 3.0
 • L x W = 3.4 x 3.0 = 10.2

 14

Rupture Generator
• Graves and Pitarka rupture generator.
 genslip_v3.3 genslip_v5.4.2

 15

Source Modelling
 Method Point Source Finite Fault 1 Finite Fault 2 Finite Fault 3
 Slip gen N/A genslip_v3.3 genslip_v5.4.2 genslip_v5.4.2
 Aspect
 Ratio N/A 1.0 1.0 ∝ L/W (Leonard)

 Slip
 One slip value

Rise Time 0.5s

 Rake One rake value

 16

Example Simulation: 2013 Lake Grassmere
 • Text
MGCS

 TFSS

 17

Waveforms
 Observed

 Point Source

 Finite Fault 3

 Observed

 Point Source

 Finite Fault 3

 18

Intensity Measures vs Rrup

PGA SA(3.0s)

 19

Analysis of Entire Dataset
• Mixed-effects regression framework to identify systematic biases.
• General form of a ground motion model for event e and station s:

 ln = + Δ

 Observation = Median Total
 +
 ln = prediction
 + residual
 (a + 2 ) + + 0

 Observation = Median Model + Systematic site + Between- Remaining
 + +
 prediction bias residual event within-event
 residual residual
 Standard deviations: 2 

 20

Model Prediction Bias

 21

Standard Deviations

 22

Future Work
• Modified Vs30 model built on the Foster et al. (2019) model.
• More investigation and analysis of 200m results.
• 100m grid run.
• Improvement of ground motion quality classification neural network.
• Subduction earthquakes.
• Uncertainty characterisation.

 23